From: Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>
To: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>,
qemu-block@nongnu.org
Cc: qemu-devel@nongnu.org, kwolf@redhat.com, mreitz@redhat.com,
stefanha@redhat.com, fam@euphon.net, jsnow@redhat.com,
armbru@redhat.com, eblake@redhat.com, den@openvz.org
Subject: Re: [PATCH v10 6/9] copy-on-read: skip non-guest reads if no copy needed
Date: Wed, 7 Oct 2020 22:28:11 +0300 [thread overview]
Message-ID: <e4153b7f-21ed-1e3e-b7e9-414bd1a4f11b@virtuozzo.com> (raw)
In-Reply-To: <de411799-a8a1-4bff-a48d-2692cc29be61@virtuozzo.com>
07.10.2020 22:01, Andrey Shinkevich wrote:
>
> On 07.10.2020 13:06, Vladimir Sementsov-Ogievskiy wrote:
>> 29.09.2020 15:38, Andrey Shinkevich wrote:
>>> If the flag BDRV_REQ_PREFETCH was set, pass it further to the
>>> COR-driver to skip unneeded reading. It can be taken into account for
>>> the COR-algorithms optimization. That check is being made during the
>>> block stream job by the moment.
>>>
>>> Signed-off-by: Andrey Shinkevich <andrey.shinkevich@virtuozzo.com>
>>> ---
>>> block/copy-on-read.c | 14 ++++++++++----
>>> block/io.c | 2 +-
>>> 2 files changed, 11 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/block/copy-on-read.c b/block/copy-on-read.c
>>> index f53f7e0..5389dca 100644
>>> --- a/block/copy-on-read.c
>>> +++ b/block/copy-on-read.c
>>> @@ -145,10 +145,16 @@ static int coroutine_fn cor_co_preadv_part(BlockDriverState *bs,
>>> }
>>> }
>>> - ret = bdrv_co_preadv_part(bs->file, offset, n, qiov, qiov_offset,
>>> - local_flags);
>>> - if (ret < 0) {
>>> - return ret;
>>> + if ((flags & BDRV_REQ_PREFETCH) &
>>
>> BDRV_REQ_PREFETCH is documented to be only used with BDRV_REQ_COPY_ON_READ. But here
>> BDRV_REQ_COPY_ON_READ appears intermediately. We should change documentation in block.h
>> in a separate patch (and probably code in bdrv_aligned_preadv())
>>
>
> OK, we will come here without the BDRV_REQ_PREFETCH flag set.
flag BDRV_REQ_PREFETCH should be set in stream job. Where should it be handled, I don't follow?
> To differ between guest reads and the stream job ones, we would set it here by checking for the qiov NULL pointer:
>
>
> diff --git a/block/copy-on-read.c b/block/copy-on-read.c
> index 4e3b1c5..df2c2ab 100644
> --- a/block/copy-on-read.c
> +++ b/block/copy-on-read.c
> @@ -144,6 +144,9 @@ static int coroutine_fn cor_co_preadv_part(BlockDriverState *bs,
> n, &n);
> if (ret) {
> local_flags |= BDRV_REQ_COPY_ON_READ;
> + if (!qiov) {
> + local_flags |= BDRV_REQ_PREFETCH;
if qiov is NULL, this means that flags must include BDRV_REQ_PREFETCH. local_flags should inherit flags I think.
> + }
> }
> }
>
> Andrey
>
>>> + !(local_flags & BDRV_REQ_COPY_ON_READ)) {
>>> + /* Skip non-guest reads if no copy needed */
>>> + } else {
>>> +
>>
>> extra new-line ?
>>
>>> + ret = bdrv_co_preadv_part(bs->file, offset, n, qiov, qiov_offset,
>>> + local_flags);
>>> + if (ret < 0) {
>>> + return ret;
>>> + }
>>> }
>>> offset += n;
>>> diff --git a/block/io.c b/block/io.c
>>> index 11df188..62b75a5 100644
>>> --- a/block/io.c
>>> +++ b/block/io.c
>>> @@ -1388,7 +1388,7 @@ static int coroutine_fn bdrv_co_do_copy_on_readv(BdrvChild *child,
>>> qemu_iovec_init_buf(&local_qiov, bounce_buffer, pnum);
>>> ret = bdrv_driver_preadv(bs, cluster_offset, pnum,
>>> - &local_qiov, 0, 0);
>>> + &local_qiov, 0, flags & BDRV_REQ_PREFETCH);
>>
>> Why? In this place we want to read. We'll write back the data a few lines below. What will we write,
>> if underlying driver decide to do nothing because of BDRV_REQ_PREFETCH?
>>
>
> See my comment above please.
Anyway, BDRV_REQ_PREFETCH here is wrong. You should not pass any qiov, if you set BDRV_REQ_PREFETCH flag.
If we come to bdrv_co_do_copy_on_readv, it means that we have COPY_ON_READ flag. And therefore, we will handle
PREFETCH and COPY_ON_READ here in generic layer. And therefore, we shouldn't pass them to driver.
On the contrary, if we have PREFETCH flag in bdrv_co_aligned_preadv, but don't have COPY_ON_READ in the same time,
this means that we must pass PREFETCH flag to the driver if it supports it. And do nothing if driver
doesn't support PREFETCH. That's how I see it.
>
>>> if (ret < 0) {
>>> goto err;
>>> }
>>>
>>
>>
--
Best regards,
Vladimir
next prev parent reply other threads:[~2020-10-07 19:29 UTC|newest]
Thread overview: 25+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-09-29 12:38 [PATCH v10 0/9] Apply COR-filter to the block-stream permanently Andrey Shinkevich via
2020-09-29 12:38 ` [PATCH v10 1/9] copy-on-read: Support preadv/pwritev_part functions Andrey Shinkevich via
2020-09-29 12:38 ` [PATCH v10 2/9] copy-on-read: add filter append/drop functions Andrey Shinkevich via
2020-10-05 13:34 ` Vladimir Sementsov-Ogievskiy
2020-10-05 16:23 ` Andrey Shinkevich
2020-09-29 12:38 ` [PATCH v10 3/9] qapi: add filter-node-name to block-stream Andrey Shinkevich via
2020-09-29 12:38 ` [PATCH v10 4/9] copy-on-read: pass base node name to COR driver Andrey Shinkevich via
2020-10-05 14:50 ` Vladimir Sementsov-Ogievskiy
2020-09-29 12:38 ` [PATCH v10 5/9] copy-on-read: limit guest COR activity to base in " Andrey Shinkevich via
2020-10-05 14:58 ` Vladimir Sementsov-Ogievskiy
2020-10-05 16:45 ` Andrey Shinkevich
2020-10-05 18:18 ` Andrey Shinkevich
2020-09-29 12:38 ` [PATCH v10 6/9] copy-on-read: skip non-guest reads if no copy needed Andrey Shinkevich via
2020-10-07 10:06 ` Vladimir Sementsov-Ogievskiy
2020-10-07 19:01 ` Andrey Shinkevich
2020-10-07 19:28 ` Vladimir Sementsov-Ogievskiy [this message]
2020-10-09 12:56 ` Andrey Shinkevich
2020-09-29 12:38 ` [PATCH v10 7/9] stream: skip filters when writing backing file name to QCOW2 header Andrey Shinkevich via
2020-10-07 10:14 ` Vladimir Sementsov-Ogievskiy
2020-09-29 12:38 ` [PATCH v10 8/9] block: remove unused backing-file name parameter Andrey Shinkevich via
2020-10-07 10:21 ` Vladimir Sementsov-Ogievskiy
2020-10-07 19:32 ` Andrey Shinkevich
2020-09-29 12:38 ` [PATCH v10 9/9] block: apply COR-filter to block-stream jobs Andrey Shinkevich via
2020-10-07 17:27 ` Andrey Shinkevich
2020-10-07 17:39 ` Andrey Shinkevich
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=e4153b7f-21ed-1e3e-b7e9-414bd1a4f11b@virtuozzo.com \
--to=vsementsov@virtuozzo.com \
--cc=andrey.shinkevich@virtuozzo.com \
--cc=armbru@redhat.com \
--cc=den@openvz.org \
--cc=eblake@redhat.com \
--cc=fam@euphon.net \
--cc=jsnow@redhat.com \
--cc=kwolf@redhat.com \
--cc=mreitz@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=stefanha@redhat.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).